ABF Dihedral

LAMMPS LAMMPS Mailing List Mirror
1

Dear Lammps users,

I am using lammps-16May18 version to do free-energy calculations using ABF method. Dihedral is my colvar. I broked my reaction coordinate into five windows and the first window spans from 0 to 46 degree, When I performing the free energy calculations I found that in .traj file the dihedral is reaching negative dihedral angles (eg: -52). To avoid this I have used hardlowerbound on but I didn’t see any difference.

As far as I know, this hardlowerbound will set to natural lower limit to the function(colvar). {eg: In case of distance lower limit is 0}

Can anyone please tell me how to restrict my simulations in specified boundaries. Moreover, In the previous versions, there is an option called lowerwallconstant which is not there in the latest version.

I want to calculate the correlation length for the series of biasing forces for that I have to get the instantaneous force on the colvar. From the documentation, I found that “Outputappliedforce on” gives the magnitude of the force. But from the simulations, I found that the force is zero. There is another option “OutputTotalForce on” which gives me non-zero forces under (fk_{name}) name. The distribution of the fk_{name} looks like Gaussian distribution. So please tell me which forces I have to take to calculate error bars.

Thanks,
Divya.

2

Hello Divya, please check the documentation for wall potentials:
http://colvars.github.io/colvars-refman-lammps/colvars-refman-lammps.html#sec:colvarbias_harmonic_walls
and for the convention used for the dihedral:
http://colvars.github.io/colvars-refman-lammps/colvars-refman-lammps.html#sec:cvc_periodic

The option lowerwallconstant has been replaced by harmonicWalls, but the old syntax is still honored so that older scripts won’t break.

You also misunderstood the meaning of hardlowerboundary (in addition to misspelling it):
http://colvars.github.io/colvars-refman-lammps/colvars-refman-lammps.html#sec:colvar_general
where it says clearly “This option does not affect simulation results, but enables some internal optimizations.”

Lastly, what do you want to compute the error bars of? Please do realize that while an arithmetic mean of a set is a well-defined quantity, measures of dispersion (i.e. error bars) assume that you have a good idea of the underlying statistical distribution of the property you are interested in. Based on that, you can choose a method to estimate this distribution, and only after you decide you can ask clarifications on the technical details about the method of your choice.

Regards,
Giacomo